This article discusses Brainerd’s rule-sampling theory of conservation learning and outlines an alternative structuralist model. It is argued that Brainerd’s theory does not account for the growth and changing structure of competence underlying the ontogenesis of conservation, because of (1) the lack of any model of the origin of the set of rules, (2) the lack of any model of recognition of applicability of rules in transfer tasks, and (3) the inability to account for interdependencies of response. An alternative structuralist model that would explain the growth and changing structure of competence is outlined. The latter model is based on the mathematical analysis of systems undergoing stagewise transitions and allows for the derivation of robust statistics in order to investigate discontinuous development. It is shown that Brainerd’s model is unsuited for such an investigation.